Football fans are about to get the ultimate view of their favorite sport.
A prototype device known as the BallCam might one day enable them to see the action from the ball's point of view, providing a different perspective when a pass sails into the hands of a receiver or a punt falls into the arms of a returner. Amazingly, it would also eliminate any confusion in the image caused by the rotation of the ball, thus giving a clean, uninterrupted view from an insider vantage point.
"It's really the right time for this," Kris Kitani, a post-doctoral fellow at Carnegie Mellon's Robotics Institute who co-developed the system, told Design News. "If you could watch a defender battling with a wide receiver and see the drama of that split-second action, it would be awesome from the spectator's point of view."
The BallCam is capable of editing out the spinning of the football and providing a usable image. (Source: Carnegie Mellon Robotics Institute)
Indeed, the BallCam would provide viewers with an experience unlike anything that's been available previously. To create that experience, Kitani worked with visiting researchers at the Robotics Institute and from the University of Electro-Communications in Tokyo. Together, they embedded a single GoPro Hero2 camera in a rubber-sheathed, plastic foam football, then wrote software to make the spinning images usable.
The software is the key, Kitani said. Because a spiraled pass from a quarterback can rotate at a surprising rate of 600 rpm, images from inside the ball would be useless without software to make sense of it all. The research team addressed that by employing techniques such as image stitching, feature extraction, and feature matching, then wrapping them together in a software package.
As a result, the BallCam is able to extract video frames that are pointing in the wrong direction, and keep those that are pointing in the correct direction. It accomplishes that by looking for the sky in every frame.
"When the camera points toward the sky, the image is really bright, and when it points down, the image is darker," Kitani told us. "So if you look at the brightness of each video frame and plot it over time, the output looks like a nice, clean sine wave."
During operation, the BallCam doesn't process the images on board. Instead, it captures 60 frames per second and then lets an off-board computer do the processing. Kitani said the team employed a MacBook Pro, but added that any notebook computer would suffice. He said the team did not write the algorithms for a parallel processing environment, but could easily do so if an application called for faster results.
To be sure, the technology faces hurdles. The camera, which is about one-third the size of an iPhone, must be covered by a clear plastic material in order to be incorporated in a football, making it impossible for it to be used officially in its current embodiment. The ability of the camera to stand up to the punishment of football is not in question, however, since the same hardware has previously been used on football helmets and spearfishing masks, as well as in the jaws of polar bears, sharks, and alligators.
For now, Kitani foresees it being employed for highlight films, training purposes, and sports science studies. It might also see use in soccer balls. "The technology is there," he said. "The cameras are durable and lightweight. You could embed multiple cameras inside a ball, transmit the images wirelessly, and create an all-new experience for spectators."
As there is no "practical application" for the entire sport of football, then it's probably the perfect fit :) Practical application is not necessary in any multi-billion dollar industry that has no practical basis in the first place.
But for any circumstance where uncontrolled changes of orientation occur, all this really is about is image stabilization, and in that sense, image stabilization is a significant issue already.
It's now standard in most video cameras.
But let's say that you had a tumbling spacecraft approaching a planet surface. You could spend millions stabilizing the craft... or you could simply let it tumble. If you wanted to image the surface on the way down and perhaps make a terrain map for later exploration... a fast camera and some software is a much cheaper and lighter way to go.
A camera in a bullet might allow the bullet be visually guided.
Anti-tank bombs that spin for stability could use it for visual targeting.
Remotely operated flying vehicles that don't require a stable body become practical (imagine rotors fixed to a cylindrical payload... no bearings or rotation)
It would be interesting to know the capture rate of the camera used. At 30 fps, the ball would be rotating about 90 to 120 degrees per frame capture, depending on the speed of the spiral. That would mean that of the 4 to 5 frames taken per revolution, only 1 or 2 of them would be pointed in a direction where they would provide usable data to be stitched together. That would probably leave an effective fps of between 7.5 to 15, definitely choppy compared to what we expect from our NFL broadcasts. However, increasing the frame rate to 120 fps, while it would certainly provide more usable frame captures per revolution, the amount of data would be greatly increased.
Just supposing that there probably is a sweet spot based on the present technology. My opinion is that the video quality is certainly good enough for now to determine what, if any, uses this might have in the future. Details about camera placement and fps will work themselves out, depending on the future demand for the product.
I would suspect that you wouldn't find many players throwing these balls into the stands anymore.
I agree, placing the camera at the end of the ball would create a better view and allow the ball to be better balanced. Then you could process the video in real-time using gyro and accelerometer data to keep the picture rotated level to the ground plane - hence smoother video without the dropped frames. Having cameras at both ends allows the ball to be thrown either direction plus adds further weight symmetry.
the camera they used seems enormous; especially considering the size of smartphone or eyewear cams.
Not sure why they mounted it on the side of the football. Seems to me it would be much better to have a cam mounted on each end - hidden in the point of the ball. Plus it would be easier to mantain the key physical characteristics of the ball.
Perhaps with the right positional sensors and algorithms, you could "unspin" the spin to view smooth tragectories toward and away from the quarterback.
Interesting article. I would really be excited about this possibility except that the NFL already has those video cameras on the wires over the field. That technology provides some great views of the action, and I wish basketball venues would go that way as well. The BallCam does offer a chance for close-up action that is unprecedented, although getting the design win could be a bit of a challenge.
For industrial control applications, or even a simple assembly line, that machine can go almost 24/7 without a break. But what happens when the task is a little more complex? That’s where the “smart” machine would come in. The smart machine is one that has some simple (or complex in some cases) processing capability to be able to adapt to changing conditions. Such machines are suited for a host of applications, including automotive, aerospace, defense, medical, computers and electronics, telecommunications, consumer goods, and so on. This radio show will show what’s possible with smart machines, and what tradeoffs need to be made to implement such a solution.